ebook2audiobook

ebook2audiobook converts non-DRM ebooks into narrated audio with chapters and metadata. It's for readers who want audio editions of legally acquired books, with a choice of speech engines and narrator voices.

• Books and scanned text: It accepts EPUB, MOBI, PDF, Word documents and plain text, with OCR for scanned pages and images; EPUB and MOBI give the best automatic chapter detection results.
• Narrator voices: Optional voice cloning uses your own recording and can clean noise or music from it, while custom trained models and fine-tuned presets provide other voice choices.
• Speech engines and languages: Supported engines include XTTSv2, Bark, Fairseq, VITS, Tacotron2, Tortoise, GlowTTS and YourTTS, with multilingual narration including English, Chinese and Arabic.
• Narration control: SML tags specify pauses and voice changes, and output formats include M4B, MP3, FLAC and WAV in mono or stereo.

The converter runs locally on Windows, macOS and Linux through a Gradio browser interface, with Docker support. It's open source under Apache 2.0. Processing can use a CPU or acceleration through CUDA, ROCm, Intel XPU, Jetson and Apple Silicon MPS. Minimum requirements are 2 GB RAM and 1 GB VRAM for GPU use; modern speech engines can be very slow on CPU. Docker doesn't expose MPS, so Apple Silicon processing in Docker uses the CPU.